U of M Spinoff Ninja Metrics Slices and Dices Social Data

by Peter Beacom

Categories

Ninja MetricsUsing proprietary machine learning software spun out from the Data Analysis and Management Group at the U of M,  Ninja Metrics aims to analyze and model social data in order to better predict online
community behavior.

As part of the Virtual World Observatory, a multi-university research collaboration, the University group has focused a large portion of initial research and analysis on EverQuest II data provided by Sony Online Entertainment (SOE); the findings of which were published in premiere data mining and social computing conferences over the last two years.

The analysis has resulted in many observations that range from the relatively obvious — such as player-versus-player combat is most popular with males ages 14-25 — to the previously unconsidered: one of the most aggressive and hardcore player segments consists of females using male avatars that self identify as lesbians. An observation valuable to SOE was that some 80% of their canceled accounts occur in early stages of game play; this identified a target market that  the game company could serve better in order improve retention rates.

Following on their success at analyzing MMOG data in an academic setting, Professor Jaideep Srivastava (University of Minnesota), Professor Dmitri Williams (University of Southern California) and Kyong Jin Shim, a PhD candidate at the University of Minnesota — officially co-founded Ninja Metrics in March of 2010. The team is actively seeking relationships with large online communities that seek to analyze and better understand archives of user behaviors. Accordingly, Jin Shim recently traveled to Silicon Valley to observe the market space and meet with prospective customers.

“Online communities may be interested in discovering a number of different behaviors. For example, not only can behavioral analysis identify which players are providing negative experience for other players, but the threshold for negative experiences that lead users to exit an online community can be determined,” says Kyong, confident that with enough behavioral data provided, Ninja Metrics can accurately predict future behavior that ultimately leads to increased revenues.

Comments

Sponsors